A new paradigm based on agents applied to free-hand sketch recognition

Abstract

devices that allow facing up to the conceptual design phase of a product. Recognizers play an important role in this field, allowing the interpretation of the user's intention, but they still present some important lacks. This paper proposes a new recognition paradigm using an agentbased architecture that does not depend on the drawing sequence and takes context information into account to help decisions. Another improvement is the absence of operation modes, that is, no button is needed to distinguish geometry from symbols or gestures, and also "interspersing"and "overtracing" are accomplished.

Improvements to the TCVD method to segment hand-drawn sketches

“…And it is this feature which makes, consequently, conceptual design stages are completely detached from the rest of the stages of design. Thus, to achieve interfaces that support natural human-computer interaction it is necessary to develop intelligent techniques for finding corners and tangent points in hand-drawn sketches and so for the automatic recognition of sketches that allows users drawing as they naturally would without any constraints, like introducing the sketches in a particular order or requiring a previous training by the user to learn a set of specified symbols or shapes [2].…”

A novel approach to 2D drawings-based reconstruction of 3D building digital models

“…Some approaches have aimed at automatically interpreting and correcting architectural drawings while they are sketched. For instance, solutions based on a multi-agents system allow to recognize graphical objects such as lines or circles but also spoken annotations [7,8]. Another approach focused on errors that can be made by an architect during the design process and highlighted the importance of detecting and correcting errors in the early phases to avoid an error propagation [9].…”